how to connect different classes?

2 views (last 30 days)
Patrick Mboma
Patrick Mboma on 26 Jul 2014
Commented: Adam on 18 Aug 2014
Dear all, Suppose I have the following function
function varargout=myfunc(f,varargin)
atmp=cell(500,4);
iter=0;
nf=numel(f);
fval=nan(nf,1);
for ii=1:nf
fval(ii)=f{ii}(varargin{:});
end
function c=plus_(a,b)
id=find(strcmp(atmp(:,1),'+') && [atmp{:,2}]==a && [atmp{:,3}]==b);
if isempty(id)
c=a+b;
iter=iter+1;
atmp(iter,:)={'+',a,b,c};
else
c=atmp{id,end};
end
end
end
This function would not work, but it serves to illustrate the question I would like to ask. The point with this function is that it has a nested function "plus_" and the variables created in the main function (atmp, iter, etc.) are also available to the nested function and the nested function can modify those variables.
In this simple example, "f" is just an array of function handles, e.g. f{1}=@(x,y)exp(x+2*y); the nested function merely adds a and b, but does the actual calculation if a+b has not been computed before. If a+b has been computed before, then the result is just fetched.
For this to work, it should be written in classes overloading operations +, -, *, /, etc. But then how to implement variables atmp and iter such that they are available to the various methods? One solution would be to define atmp and iter as global variables to each of the overloaded methods and initialize it outside the class. A better solution, perhaps, would use handles, notify, events, addlistener, etc. but I do not know how to implement them. Any ideas?
Thanks, P.
  5 Comments
Patrick Mboma
Patrick Mboma on 28 Jul 2014
Per,
Thanks again for the links. I will look into them although I don't feel like learning python just yet.
Adam
Adam on 18 Aug 2014
I'm confused what the question is here. Your title and the actual part of your post with a ? talks about classes yet everything else seems to be related to nested functions.
If you have a class that overloads the + operator and any other operators then properties of the class are available to the overloaded operators in the nature of being a class.

Sign in to comment.

Answers (1)

Aurele Turnes
Aurele Turnes on 4 Aug 2014
Edited: Aurele Turnes on 4 Aug 2014
Like per isakson pointed out, I think that you can achieve this by using memoization techniques. The following page on MATLAB Central gives a pretty good example of how to do this:
In order to keep the same types of variable that you are using, your function could look something like this:
function f = memoize()
%one-arg F, inputs testable with ==
%allow nonscalar input.
atmp = cell(500,4);
iter = 0;
f = @inner;
function out = inner(a,b,F)
% find which in's already computed in x
if iter>0
ind= strcmp(atmp(1:iter,1),'+') & (cell2mat(atmp(1:iter,2))==a) & (cell2mat(atmp(1:iter,3))==b);
if any(ind)
out = atmp{ind,4};
else
out = eval(['a' F 'b']);
iter = iter+1;
atmp(iter,:) = {F, a , b, out};
end
else
iter = iter+1;
out = eval(['a' F 'b']);
atmp(iter,:) = {F, a , b, out};
end
end
end
You can then call it and use it as follows:
f = memoize
f(3,4,'+')
f(3,5,'+')
f(3,4,'+')
f(3,4,'*')
Look at the example function memoize2 in the link above to vectorize this technique.
  1 Comment
Patrick Mboma
Patrick Mboma on 18 Aug 2014
Dear Aurele, I think I fairly well understand the concept of memoization and I use it in other applications. But for this one, my problem remains and comes from the fact that the various inputs are objects of a certain class which overloads the operations +, -, *, /, ^, cos, log, exp, etc.
Thanks

Sign in to comment.

Categories

Find more on Performance and Memory in Help Center and File Exchange

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!